Processes, machines, and compositions related to analyzing neoplasms such as cancer

Abstract

Analyzing whether an individual has a neoplasm, such as an early-stage cancer, is based on detection of novel cancer-associated biomarkers in a biological sample. In one aspect, a process begins with a liquid biopsy from the individual, e.g., a blood sample, which is subjected to several sample processing steps resulting in a processed sample, such as a next generation sequencing (NGS) sample library. The processed sample contains cancer-informative cell-free DNA (“cfDNA”) from the liquid biopsy. The processed sample is subjected to DNA sequencing, to detect certain low-abundance cfDNA, for example by using a next generation DNA sequencer. Sequencer data output from the DNA sequencer is processed by a data processing system to provide a classification for the individual. For example, the data processing system can determine a likelihood of the individual having a neoplasm such as an early-stage cancerous tumor based on that individual's liquid biopsy.

What is claimed is: 
     
         1 . A process for detecting a risk of presence of an early stage neoplasm in a subject based on a liquid biopsy sample from the subject, the liquid biopsy sample containing cell-free DNA fragments, the process comprising:
 determining methylation information of the cell-free DNA fragments originating from cancer informative CGIs;   using an analytical platform, computing, for one or more of the cancer-informative CGIs, a respective plurality of metrics based on methylation information of respective located cell-free DNA fragments originating from the cancer-informative CGI, wherein the plurality of metrics comprises at least a transition metric reflecting a number of transitions between differentially methylated neighboring CpG sites;   processing the plurality of metrics for the cancer-informative CGIs with a computational model, the computational model providing an output indicating a likelihood of presence in the individual of an early stage neoplasm, wherein the computational model is based on a training set including samples containing the cancer informative CGIs originating from individuals known to have the early stage neoplasm.   
     
     
         2 . The process of  claim 1 , further comprising detecting or monitoring, by the analytical platform, presence of early-stage cancer or a precancerous state or condition or to predict a likelihood of a cancer or a precancerous state or condition using the cell-free DNA fragments. 
     
     
         3 . The process of  claim 1 , wherein the plurality of metrics comprises at least a proportion metric. 
     
     
         4 . The process of  claim 1 , wherein the liquid biopsy sample comprises plasma obtained from an asymptomatic individual. 
     
     
         5 . The process of  claim 1 , further comprising predicting, based on one or more metrics of the plurality of metrics and the cancer informative CGIs, a likelihood of presence of a neoplasm in the subject. 
     
     
         6 . The process of  claim 5 , wherein the neoplasm is an early-stage cancerous solid tumor. 
     
     
         7 . The process of  claim 1 , wherein the selected cancer informative CGIs include those having a statistical metric above a threshold based on a training set and one or more metrics computed for samples in the training set. 
     
     
         8 . The process of  claim 7 , wherein the one or more metrics comprises at least a proportion metric. 
     
     
         9 . The process of  claim 1 , wherein the computational model predicts the likelihood of presence of a precancerous neoplasm or early stage cancer in the subject with a sensitivity greater than a first threshold and a specificity greater than a second threshold. 
     
     
         10 . The process of  claim 9 , the sensitivity is greater than about 70% at a specificity of greater than about 95%. 
     
     
         11 . The process of  claim 1 , wherein the cancer informative CGIs are selected from a group consisting of a ranked set of candidate CGIs selected from TABLE A. 
     
     
         12 . The process of  claim 1 , wherein the average number of cell-free DNA fragments processed per cancer informative CGI is greater than 200. 
     
     
         13 . A machine for detecting a risk of presence of an early stage neoplasm in a subject based on a liquid biopsy sample from the subject, the liquid biopsy sample containing cell-free DNA fragments, the machine comprising:
 computer storage which stores data representing methylation information of cell-free DNA fragments from the liquid biopsy sample and originating from cancer informative CG islands (CGIs)   a processing system accessing the computer storage to compute, for one or more cancer-informative CGI, a respective plurality of metrics based on the methylation information of the respective located cell-free DNA fragments originating from the cancer-informative CGI, and to store for the subject, for each of the cancer informative CGIs, the respective computed plurality of metrics for the cancer informative CGI, wherein the plurality of metrics comprises a transition metric reflecting a number of transitions between differentially methylated neighboring CpG sites; and   the processing system further processing the plurality of metrics for the cancer-informative CGIs with a computational model, the computational model providing an output indicating a likelihood of presence in the individual of an early stage neoplasm, wherein the computational model is based on a training set including samples containing the cancer informative CGIs originating from individuals known to have the early stage neoplasm.   
     
     
         14 . The machine of  claim 13 , wherein the average number of cell-free DNA fragments processed per cancer informative CGI is greater than 200. 
     
     
         15 . The machine of  claim 13 , wherein the plurality of metrics comprises at least a proportion metric. 
     
     
         16 . The machine of  claim 13 , wherein the computational model is configured to predict a likelihood of presence of a neoplasm in the subject based on one or more metrics of the computed plurality of metrics and the cancer informative CGIs. 
     
     
         17 . The machine of  claim 16 , wherein the neoplasm is an early-stage cancerous solid tumor. 
     
     
         18 . The machine of  claim 13 , wherein the selected cancer informative CGIs include those having a statistical metric above a threshold based on a training set and one or more metrics computed for samples in the training set. 
     
     
         19 . The process of  claim 18 , wherein the one or more metrics comprises at least a proportion metric. 
     
     
         20 . The machine of  claim 13 , wherein the average number of cell-free DNA fragments processed per cancer informative CGI is greater than 200.